[expand docs a bit marijnh@gmail.com**20080605083659] { hunk ./doc/index.html 15 - simplify some kinds of parallelisation. It uses a thread pool to - schedule small computations without allocating a thread for them. - This makes it possible to make use of multiple cores without too - much fuss.

+ simplify 'result-oriented' parallelism. It uses a thread pool to + concurrently run small computations without spawning a new thread. + This makes it possible to exploit multiple cores without much + extra fuss.

hunk ./doc/index.html 83 - existence you should be aware of. + existence you should be aware of. It starts inactive, but as + soon as a task is created, it is initialised. See thread-pool-size and + finish-tasks for ways + in which it can be managed. + hunk ./doc/index.html 92 - that may be parallelised. + that may be parallelised, and enqueues it. When given the + chance, the threads in the thread pool will take such a task, + execute it, and store its result. + hunk ./doc/index.html 97 - extracts the value from a computation. + extracts the value from a computation. This blocks when the task + is still running. If it has not yet started, the joining thread + will claim and execute the task itself. hunk ./doc/index.html 116 - to use more hardware -- or make better use of the hardware we + to use more hardware ― or make better use of the hardware we hunk ./doc/index.html 124 -

On my 2-core machine, that speeds things up by about a third -- +

On my 2-core machine, that speeds things up by about a third ― hunk ./doc/index.html 249 - be called when shutting down -- killing the threads might cause - some tasks to be aborted, which could result in data loss. If you - join every task you create, this should not be necessary. Note - that if you call this function while other threads are still + be called when shutting down ― killing the threads might + cause some tasks to be aborted, which could result in data loss. + If you join every task you create, this should not be necessary. + Note that if you call this function while other threads are still }